grida-research/typebox

JSON Schema Type Builder with Static Type Resolution for TypeScript

TypeBox

JSON Schema Type Builder with Static Type Resolution for TypeScript

Install

$ npm install @sinclair/typebox --save

Overview

TypeBox is a type builder library that allows developers to compose in-memory JSON Schema objects that can be statically resolved to TypeScript types. The schemas produced by TypeBox can be used directly as validation schemas or reflected upon by navigating the standard JSON Schema properties at runtime. TypeBox can be used as a simple tool to build up complex schemas or integrated into RPC or REST services to help validate JSON data received over the wire.

TypeBox does not provide any mechanism for validating JSON Schema. Please refer to libraries such as AJV or similar to validate the schemas created with this library.

Requires TypeScript 4.0.3 and above.

License MIT

Contents

• Install
• Overview
• Example
• Types
• Function Types
• Functions
• Generics
• Validation

Example

The following shows the general usage.

import { Type, Static } from '@sinclair/typebox'

// some type ...

type Order = {
    email:    string,
    address:  string,
    quantity: number,
    option:   'pizza' | 'salad' | 'pie'
}

// ... can be expressed as ...

const Order = Type.Object({
    email:    Type.String({ format: 'email' }), 
    address:  Type.String(),
    quantity: Type.Number({ minimum: 1, maximum: 99 }),
    option:   Type.Union([
        Type.Literal('pizza'), 
        Type.Literal('salad'),
        Type.Literal('pie')
    ])
})

// ... which can be reflected

console.log(JSON.stringify(Order, null, 2))

// ... and statically resolved

type TOrder = Static<typeof Order>

// .. and validated as JSON Schema

JSON.validate(Order, {  // IETF | TC39 ?
    email: '[email protected]', 
    address: '...', 
    quantity: 99, 
    option: 'pie' 
}) 

// ... and so on ...

Types

TypeBox provides a number of functions to generate JSON Schema data types. The following tables list the functions TypeBox provides and their respective TypeScript and JSON Schema equivalents.

TypeBox > TypeScript

Type	TypeBox	TypeScript
Literal	const T = Type.Literal(123)	type T = 123
String	const T = Type.String()	type T = string
Number	const T = Type.Number()	type T = number
Integer	const T = Type.Integer()	type T = number
Boolean	const T = Type.Boolean()	type T = boolean
Object	const T = Type.Object({ name: Type.String() })	type T = { name: string }
Array	const T = Type.Array(Type.Number())	type T = number[]
Map	const T = Type.Map(Type.Number())	type T = { [key: string] } : number
Intersect	const T = Type.Intersect([Type.String(), Type.Number()])	type T = string & number
Union	const T = Type.Union([Type.String(), Type.Number()])	type T = string | number
Tuple	const T = Type.Tuple([Type.String(), Type.Number()])	type T = [string, number]
Any	const T = Type.Any()	type T = any
Null	const T = Type.Null()	type T = null
Pattern	const T = Type.Pattern(/foo/)	type T = string
Guid	const T = Type.Guid()	type T = string

TypeBox > JSON Schema

Type	TypeBox	JSON Schema
Literal	const T = Type.Literal(123)	{ type: 'number', enum: [123] }
String	const T = Type.String()	{ type: 'string' }
Number	const T = Type.Number()	{ type: 'number' }
Integer	const T = Type.Number()	{ type: 'integer' }
Boolean	const T = Type.Boolean()	{ type: 'boolean' }
Object	const T = Type.Object({ name: Type: String() })	{ type: 'object': properties: { name: { type: 'string' } }, required: ['name'] }
Array	const T = Type.Array(Type.String())	{ type: 'array': items: { type: 'string' } }
Map	const T = Type.Map(Type.Number())	{ type: 'object', additionalProperties: { type: 'number' } }
Intersect	const T = Type.Intersect([Type.Number(), Type.String()])	{ allOf: [{ type: 'number'}, {type: 'string'}] }
Union	const T = Type.Union([Type.Number(), Type.String()])	{ anyOf: [{ type: 'number'}, {type: 'string'}] }
Tuple	const T = Type.Tuple([Type.Number(), Type.String()])	{ type: "array", items: [{type: 'string'}, {type: 'number'}], additionalItems: false, minItems: 2, maxItems: 2 }
Any	const T = Type.Any()	{ }
Null	const T = Type.Null()	{ type: 'null' }
Pattern	const T = Type.Pattern(/foo/)	{ type: 'string', pattern: 'foo' }
Guid	const T = Type.Guid()	{ type: 'string', pattern: '<guid-regex>' }

Type Modifiers

The following are object property modifiers. Note that Type.Optional(...) will make the schema object property optional. Type.Readonly(...) however has no effect on the underlying schema as is only meaningful to TypeScript.

Type	TypeBox	TypeScript
Readonly	const T = Type.Object({ email: Type.Readonly(Type.String()) })	type T = { readonly email: string }
Optional	const T = Type.Object({ email: Type.Optional(Type.String()) })	type T = { email?: string }

Enums

It is possible to define TypeScript enums and use them as part of your TypeBox schema. Both number and string-valued enums are supported.

enum Color {
    Red = 'red',
    Blue = 'blue'
}

const T = Type.Enum(Color); // -> json-schema: `{ enum: ['red','green'] }`

Note that the generated json-schema will only permit the values of the enum, not its keys. In TypeScript, if you omit the value for an enum option, TypeScript will implicitly assign the option a numeric value.

E.g.:

enum Color {
    Red, // implicitly gets value `0`
    Blue // implicitly gets value `1`
}

const T = Type.Enum(Color); // -> json-schema: `{ enum: [0, 1] }`

User Defined Schema Properties

It's possible to specify custom properties on schemas. The last parameter on each TypeBox function accepts an optional UserDefinedOptions object. Properties specified in this object will appear as properties on the resulting schema object. Consider the following.

const T = Type.Object({
    value: Type.String({ 
        description: 'A required string.'
    })
}, {
    description: 'An object with a value'
})

{
  "description": "An object with a value",
  "type": "object",
  "properties": {
    "value": {
      "description": "A required string.",
      "type": "string"
    }
  },
  "required": [
    "value"
  ]
}

Function Types

TypeBox allows function signatures to be composed in a similar way to other types, but uses a custom schema format to achieve this. Note, this format is not JSON Schema, rather it embeds JSON Schema to encode function arguments and return types. The format also provides additional types not present in JSON Schema; Type.Constructor(), Type.Void(), Type.Undefined(), and Type.Promise().

For more information on using functions, see the Functions and Generics sections below.

Format

The following is an example of how TypeBox encodes function signatures.

type T = (a: string, b: number) => boolean

{
    "type": "function",
    "returns": { "type": "boolean" },
    "arguments": [
        {"type": "string" }, 
        {"type": "number" },
    ]
}

TypeBox > TypeScript

Intrinsic	TypeBox	TypeScript
Function	const T = Type.Function([Type.String()], Type.String())	type T = (arg0: string) => string
Constructor	const T = Type.Constructor([Type.String()], Type.String())	type T = new (arg0: string) => string
Promise	const T = Type.Promise(Type.String())	type T = Promise<string>
Undefined	const T = Type.Undefined()	type T = undefined
Void	const T = Type.Void()	type T = void

TypeBox > JSON Function

Intrinsic	TypeBox	JSON Function
Function	const T = Type.Function([Type.String()], Type.Number())	{ type: 'function', arguments: [ { type: 'string' } ], returns: { type: 'number' } }
Constructor	const T = Type.Constructor([Type.String()], Type.Number())	{ type: 'constructor', arguments: [ { type: 'string' } ], returns: { type: 'number' } }
Promise	const T = Type.Promise(Type.String())	{ type: 'promise', item: { type: 'string' } }
Undefined	const T = Type.Undefined()	{ type: 'undefined' }
Void	const T = Type.Void()	{ type: 'void' }

Functions

The following demonstrates creating function signatures for the following TypeScript types.

TypeScript

type T0 = (a0: number, a1: string) => boolean;

type T1 = (a0: string, a1: () => string) => void;

type T2 = (a0: string) => Promise<number>;

type T3 = () => () => string;

type T4 = new () => string

TypeBox

const T0 = Type.Function([Type.Number(), Type.String()], Type.Boolean())

const T1 = Type.Function([Type.String(), Type.Function([], Type.String())], Type.Void())

const T2 = Type.Function([Type.String()], Type.Promise(Type.Number()))

const T3 = Type.Function([], Type.Function([], Type.String()))

const T4 = Type.Constructor([], Type.String())

Generics

Generic function signatures can be composed with TypeScript functions with Generic Constraints.

TypeScript

type ToString = <T>(t: T) => string

TypeBox

import { Type, Static, TStatic } from '@sinclair/typebox'

const ToString = <G extends TStatic>(T: G) => Type.Function([T], Type.String())

However, it's not possible to statically infer what type ToString is without first creating some specialized variant of it. The following creates a specialization called NumberToString.

const NumberToString = ToString(Type.Number())

type X = Static<typeof NumberToString>

// X is (arg0: number) => string

To take things a bit further, the following code contains some generic TypeScript REST setup with controllers that take some generic resource of type T. Below this we express that same setup using TypeBox. The resulting type IRecordController contains reflectable interface metadata about the RecordController.

TypeScript

interface IController<T> {
    get    (): Promise<T>
    post   (resource: T): Promise<void>
    put    (resource: T): Promise<void>
    delete (resource: T): Promise<void>
}

interface Record {
     key: string
     value: string
}

class RecordController implements IController<Record> {
    async get   (): Promise<Record> { throw 'not implemented' }
    async post  (resource: Record): Promise<void> { /* */  }
    async put   (resource: Record): Promise<void> { /* */  }
    async delete(resource: Record): Promise<void> { /* */  }
}

TypeBox

import { Type, Static, TStatic } from '@sinclair/typebox'

const IController = <G extends TStatic>(T: G) => Type.Object({
    get:    Type.Function([], Type.Promise(T)),
    post:   Type.Function([T], Type.Promise(Type.Void())),
    put:    Type.Function([T], Type.Promise(Type.Void())),
    delete: Type.Function([T], Type.Promise(Type.Void())),
})

type Record = Static<typeof Record>
const Record = Type.Object({
    key: Type.String(),
    value: Type.String()
})

type IRecordController = Static<typeof IRecordController>
const IRecordController = IController(Record)

class RecordController implements IRecordController {
    async get   (): Promise<Record> { throw 'not implemented' }
    async post  (resource: Record): Promise<void> { /* */  }
    async put   (resource: Record): Promise<void> { /* */  }
    async delete(resource: Record): Promise<void> { /* */  }
}

// Reflect
console.log(IRecordController)

Validation

The following uses the library Ajv to validate a type.

import * Ajv from 'ajv'

const ajv = new Ajv({ })

ajv.validate(Type.String(), 'hello')  // true

ajv.validate(Type.String(), 123)      // false